Compare
Specialist vs LLM
Same task, two systems, quality, latency and £ on a frozen set.
NLP Advanced Techniques
Beyond “call the LLM”: tokenisation, sequence labelling, retrieval text, evaluation, and when a specialist model still beats a general one.
Created by Baljeet Dogra
Course Objectives
- Debug a tokenisation failure instead of blaming the model.
- Train or adapt a specialist model and compare it to a general LLM on cost and quality.
- Evaluate generation and classification with the right metrics.
- Handle multilingual and messy text without pretending English is the world.
What you will produce
Lab
Labelling pipeline
NER or classification with error analysis, not a leaderboard screenshot.
6 weeks curriculum
Expand a part for the syllabus. Content stays searchable when closed.
01 Text as tokens Week 1
Tokenisers, multilingual penalties, off-by-one structured output.
- Tokenisers
- Multilingual
- Truncation
- Offsets
02 Classify and label Weeks 2–3
Classifiers, NER, span labelling, class imbalance.
- Classification
- NER
- Imbalance
- Error analysis
03 Retrieval text Week 4
Chunking, queries, hybrid features — the NLP under RAG.
- Chunking
- Queries
- Hybrid
- Eval
04 Generate and compare Weeks 5–6
Metrics for generation, specialist vs LLM, capstone report.
- Generation metrics
- Cost
- Specialist models
- Capstone
Who this is for
Engineers in language products
Search, support, documents. An LLM is not automatically the system.
ML practitioners
You want Hugging Face without cargo-cult pipelines.
Prerequisites
- Python
- Deep Learning Specialisation or equivalent
- Comfort reading papers at a high level
Not a fit if
- Prompt-only ChatGPT courses
- A first course in Python
Related: Deep Learning Specialisation · Production Generative AI Systems
Questions
Is this transformers from scratch?
Working depth, not a from-scratch GPT. Deep Learning Specialisation covers more of the internals.
Hugging Face only?
It is the lab stack. You should be able to leave it.
Ready to start?
Six weeks to NLP you can evaluate against a general LLM. Create an account to enrol.
Enrol now